VALIDATION OF COUPLED NEUTRONICS / THERMAL HYDRAULICS CODES FOR VVER REACTORS (VALCO)

Frank-Peter Weiss1, Siegfried Mittag1, Siegfried Langenbuch2, Timo Vanttola3, Anitta Hämäläinen3, András Keresztúri4, Jan Hádek5, Petr Darilek6, Petko Petkov7, Alexan-

der Kuchin8, Peter Hlbocky9, Dusan Sico10, Serge Danilin11, and David Powney12

1. Introduction

Modern safety standards require the modelling of complex transients where there is a strong interaction between the thermal-hydraulic system behaviour and space-dependent neutron ki-netics. Therefore, the project is aimed at the improvement of the validation of coupled neu-tronics / thermal-hydraulics codes especially for VVER reactors. These coupled codes need to be validated against well-specified NPP transient scenarios. The EU FP5 VALCO project is partially based on results earlier obtained within the EU Phare project (SRR1/95) (Ref. [1,2]). Two selected transients, one for each a VVER-440 and a VVER-1000, were analysed in this former project by different coupled-code systems. The calculated results were compared with measured transient data from original NPPs. However, in both transients, there were significant differences in the calculated reactor power, which were suspected to be due to differences in the control rod efficiencies and in the dynamic thermal physics of the fuel rod model that affects the Doppler feedback. To study neutronics effects separately, it was therefore recommended to validate the neutronics codes including the nuclear cross section data against measurements in zero-power reactors. Furthermore, in recent years recommendations have been given for safety analyses to com-plement best estimate calculations by quantitative uncertainty analyses. Uncertainty analysis is needed to enable useful conclusions from best-estimate codes; otherwise single values of unknown accuracy would be used for comparison with certain limits of acceptance. Based on these former conclusions, the primary work programme of the current project has been derived.

2. Work programme

While the transients analysed in the Phare SRR1/95 project were initiated by perturbations in the secondary circuit, transients caused by actions in the primary circuit are of special interest to the current project, e.g. transients initiated by switching-off main coolant pumps. The pur-

1 Forschungszentrum Rossendorf e.V., FZR (D) 2 Gesellschaft für Anlagen- und Reaktorsicherheit (GRS) mbH (D) 3 Technical Research Centre of Finland, VTT (FIN) 4 KFKI Atomic Energy Research Institute, AEKI (HU) 5 Nuclear Research Institute Rez, plc, NRI (CZ) 6 VUJE Trnava a.s. (SK) 7 Institute for Nuclear Research and Nuclear Energy, INRNE (BG) 8 State Scientific and Technical Centre on Nuclear and Radiation Safety, SSTCNRS (UA) 9 SE, a.s.EBO, o.z., Jaslovské Bohunice (SK) 10 SE, a.s.EBO, o.z., Mochovce (SK) 11 Russian Research Center “Kurchatov Institute”, KI (RU) 12 Serco Assurance, Winfrith (UK)

54

pose of Work Package 1, led by VTT, is therefore to extend and qualify the measurement data base for the validation of coupled codes. The analyses of new transients are to be performed with the following coupled codes: DYN3D-ATHLET, KIKO3D-ATHLET, BIPR8-ATHLET and HEXTRAN-SMABRE. The previous transient analyses have shown that the results of calculations depend on various input parameters of the codes, model options, nodalisation etc. On the one hand, different macroscopic cross section libraries and other physical model parameters have caused devia-tions between the different code options. On the other hand, differences in the results of tran-sient analyses were observed, when calculations were performed by using the same code sys-tem and input deck, but by different users. These findings initiated to adapt and apply an un-certainty analysis method to coupled codes, which is the main objective of Work Package 2, under the leadership of GRS. The first step of the uncertainty analysis is to identify and quan-tify all potentially important input parameters including their uncertainty bands and probabil-ity distributions. The propagation of the input uncertainties through the code runs provides the related probability distributions for the code results. To separate the pure neutron kinetics effects from feedback effects, a specific validation of neutron kinetics (”neutronics”) models is to be performed in Work Package 3, led by FZR, by calculations of kinetics experiments, carried out in the V-1000 zero power test facility of the Kurchatov Institute Moscow. A broad spectrum of measurements is available. The V-1000 data are considered a unique material for the validation of neutron-kinetic codes for hexago-nal fuel-assembly geometry. In a first validation step, measured V-1000 steady-state power distributions can be used to validate the three-dimensional two-group diffusion models, which form the ”stationary kernels” of the respective neutron-kinetics (dynamic) codes applied in the transient calculations. Results of two transient experiments carried out in the V-1000 zero power test facility have to be made available, in which different control rods were moved. These transients are to be calculated by the three-dimensional neutron kinetics codes DYN3D, HEXTRAN and KIKO3D. Libraries of two-group diffusion and kinetics parameters have to be generated, as a part of the project, by multi-group transport cell codes for the V-1000 fuel assemblies as well as for the radial and axial reflectors of the core.

3. Main achievements

3.1. Work package 1: Validation of coupled codes against measured VVER transients

Five new VVER transients were documented and added to the data base, that can be used for the validation of coupled neutronics / thermal-hydraulics codes. Coupled code validation was done with the different code combinations for the following two transients: • Drop of a control rod at nominal power at Bohunice-3 (VVER-440), • Coast-down of 1 from 3 MCPs at Kozloduy-6 (VVER –1000). The Bohunice case was a real and unintended plant transient. The Kozloduy transient was part of the plant start-up tests. A short characteristic of both transients is given in Table I. In the Bohunice measurement, the behaviour of the core during the rod drop and during the three minutes after starting control actions to reach 85 % power level is the most essential part

55

Table 1: Basic information of transients selected for code validation Plant VVER 440/213,

Bohunice 3, Slovakia VVER-1000/V-320, Kozloduy 6, Bulgaria

Initial reactor state 100 %, nominal state

65 %, three of four MCPs running

Transient type Control rod drop and power control action (unintended event)

Coast-down of a second MCP (start up test)

Calculation aspects Asymmetry in core Asymmetry between loops

0 200 400 600 800 1000Time (s)

24.00

26.00

28.00

30.00

Hot

-col

d le

g te

mp.

diff

. (de

g.C

)

MeasurementHEXTRAN-SMABRE (VTT)DYN3D-RELAP5 (VUJE)DYN3D-ATHLET (NRI)KIKO3D-ATHLET (AEKI)BIPR8-ATHLET (KI)

Fig. 1: Calculated and measured average coolant heat-up in the Bohunice VVER-440 core

for validation. The general behaviour during the transients was quite well calculated with all the codes: Both the power distribution changes and the fuel assembly outlet temperatures were mostly well reproduced. As an example, the coolant heat-up, which is a measure of the reactor power, is depicted in Figure 1. All the calculated hot leg – cold leg temperature differences are within a 1.5°C band around the measurements. Simultaneous measurements of the individual assembly outlet temperatures and the hot leg temperatures indicate that the coolant mixing in the upper plenum was weak during the tran-sient. That was demonstrated by those codes that included a mixing model and a detailed enough core channel model. Differing results were observed in the axial power profile, in the calculated control rod worth, and in the fuel temperatures. The differences were large in the required control rod group movement to reach the same final power. The features, that make the Kozloduy transient interesting, such as the lowered power and reverse flow in the stopped loop in the initial state, also proved to be difficult both for data collection and for modelling. The experimental data also include a record from full power nominal state, which in principle enables the definition of the initial state for the calculations. Some more specific data would, however, have been needed about the pump coast down

56

characteristics and the control regime. Anyway, the general behaviour of the Kozloduy sec-ond pump trip was reproduced by all the codes.

3.2. Work package 2: Uncertainty analysis

In a first step, the comparison of results for the Loviisa transient “One turbo-generator load drop” [2] obtained by two coupled codes using 3D-neutronics, namely SMABRE-HEXTRAN (VTT) and ATHLET-DYN3D (FZR), and the results from ATHLET using point kinetics re-vealed good consistency in view of relevant sensitivities of input parameters. The final evaluation will further take into account results from ATHLET-DYN3D (NRI) and ATHLET-KIKO3D (KFKI). The sensitivities of the code results against the variations of the input pa-rameters were evaluated using the statistical code package SUSA (Ref. [3]) of GRS.

200 400 600 800Time (s)

280.0

290.0

300.0

Hot

-leg

tem

pera

ture

(deg

.C)

MeasurementCALCULATIONUpper tolerance limitLower tolerance limit

Fig. 2: Measurement and calculation for the hot-leg temperature in the SRR1/95 Loviisa VVER-440 transient. SUSA-calculated tolerance interval

Twelve input parameters have been identified in the coupled-code calculations as the main sources of uncertainty in the Loviisa (VVER-440) and Balakovo (VVER-1000) transients ear-lier studied in the Phare SRR1/95 project. The uncertainty analysis of the Loviisa transient, e.g., led to the conclusion that the following three parameters are mainly responsible for un-certainties in the calculations: the secondary-circuit pressure, the control-assembly position (as a function of time), and the control-assembly efficiency. “Uncertainty bands” were derived for safety-relevant output parameters, like power, coolant heat-up in the core, primary pressure, mass flow, and pressurizer level. An example for the VVER-440 is given in Figure 2. The SUSA tolerance interval contains at least 90% of the un-certainty with a confidence level of at least 95%. The interval fully covers the measurement.

3.3. Work package 3: Validation of neutron-kinetic models

All neutron kinetic codes have well reproduced the significant radial tilt observed in both measured V-1000 steady-state power distributions. The good description is mostly due to the accuracy of the HELIOS / MARIKO model applied to calculate the asymmetric radial reflec-tor constants. The initially observed power deviations of up to 30% between calculation and measurement had led to the re-measuring of special V-1000 geometrical parameters, such as

57

the water-gap widths between the core and the radial reflector. The more precise geometric information then resulted in a clear reduction of the maximum power deviation down to 10%. Inner-assembly pin-power distributions, have been calculated by two neutronic codes. The maximum deviation between measurement and both the fine-mesh diffusion code HEX2DB (INRNE) and the nodal diffusion code DYN3D, applying pin-power recovery, is below 5% in the un-rodded case. When the control rods of group X are inserted in the considered assem-bly, introducing strong heterogeneity, the maximum deviation is about 9%. Systematic over-estimations of the effective multiplication factors, that were calculated by all codes for the investigated V-1000 steady states, seem to be mainly due to uncertainties in the nuclear input data of the cell codes and to the measured critical boron concentration, which was also input in the cell calculations.

0 500 1000 1500Time (s)

0.01

0.10

1.00

Rel

ativ

e po

wer

MeasurementKIKO3D (AEKI)HEXTRAN (VTT)DYN3D (FZR)

Fig. 3: V-1000 transient: Relative power versus time in FA 86 at the height of 175 cm

The two measured transients have been successfully calculated by the kinetics codes in-volved. A good agreement between the calculated and measured relative powers as functions of time has been reached, as Figure 3 shows for one example: Starting from a V-1000 steady state with all control rods withdrawn, a transient was initiated by inserting a single cluster of group IX and then withdrawing it again. The induced change of the local power density has been measured and calculated in some fuel assemblies by micro fission chambers in certain height positions.

4. Conclusions

The instrumentation of operating nuclear power plants is not primarily designed for measur-ing data suitable for coupled-code validation. Nevertheless, available information of several VVER transients has been gathered and carefully documented. In the end, the comparison be-tween different codes and the validation against the selected NPP transients was successful. The calculations were performed without additional corrections or special “tuning”.

58

The SUSA uncertainty-analysis method has been successfully extended to coupled-code cal-culations that include 3D neutron kinetics. SUSA analyses for VVER transients have contrib-uted to explain deviations between calculations and measurements, so far at least for the Phare SRR1/95 transients. The Bohunice-transient calculations, as well as the uncertainty analysis for the former Loviisa transient point to the fact that results are very sensitive to the treatment of VVER-440 control assemblies. They are quite different from control rods in Western PWR and need further at-tention in modelling by both cell codes and nodal kinetics codes. The validation against measurements in the V-1000 facility has demonstrated that the neu-tronics codes, including the applied nuclear-data input, are suitable for the calculation of power distributions and power changes caused by control-rod movements in a VVER-1000. By the same token, it would be very useful to validate VVER-440 control-assembly models, against measurements in a suitable zero-power facility. In principle, the LR0 facility of NRI Rez will fit for this purpose. Both the VALCO project and its predecessor Phare SRR1/95 have contributed to the im-proved validation state of codes for VVER reactors. Transient measurements in Western NPP could enable an improvement in the coupled-code validation for these reactor types, too.

References

[1] S. Mittag, S. Kliem, F.-P. Weiß, R. Kyrki-Rajamäki, A. Hämäläinen, S. Langenbuch, S. Danilin, J. Hadek, G. Hegyi, A. Kuchin, and D. Panayotov (2001), Validation of Coupled Neutron Kinetic / Thermal-Hydraulic Codes. Part 1: Analysis of a VVER-1000 Transient (Balakovo-4), Annals of Nuclear Energy, 28, 857

[2] Hämäläinen, R. Kyrki-Rajamäki, S. Mittag, S. Kliem, F.-P. Weiß, S. Langenbuch, S. Danilin, J. Hadek, and G. Hegyi (2002), Validation of Coupled Neutron-Kinetic / Thermal-Hydraulic Codes. Part 2: Analysis of a VVER-440 Transient (Loviisa-1), Annals of Nuclear Energy, 29, 255

[3] M. Kloos and E. Hofer (2002), SUSA: User’s Guide and Tutorial, GRS-report

59